1 Do Political Budget Cycles Differ in Latin American Democracies? Lorena G. Barberia George Avelino EconomÃ-a, Volume 11, Number 2, Spring 2011, pp (Article) Published by Brookings Institution Press DOI: /eco For additional information about this article Access Provided by Universidade de SÃ o Paulo at 11/28/12 7:02PM GMT
2 LORENA G. BARBERIA GEORGE AVELINO Do Political Budget Cycles Differ in Latin American Democracies? The literature on political business cycles has produced important insights on the extent to which politicians attempt to manipulate government monetary and fiscal policies to influence electoral outcomes. In particular, some of the strongest evidence produced to date suggests that electoral cycles are particularly marked in the case of government expenditures. 1 Governments in developing countries and so-called new democracies are often considered to be the most susceptible to the manipulation of fiscal and monetary policy to enhance their chances of reelection. 2 The experiences of recently reestablished Latin American democracies in a period marked by episodes of heightened macroeconomic volatility followed by the adoption of painful stabilization measures provide fertile ground for testing political budget cycle theories and recently formulated arguments on the acuteness of these patterns for young fragile democratic regimes in developing countries. The case of Argentina is illustrative of the expected trajectory in Latin American countries. Following the return of democracy in 1983, the Alfonsín and Menem administrations spent an average of percent of GDP and collected an average of percent of GDP in tax revenue. During this early period of democracy, fiscal deficits averaged 2.28 percent of GDP. Budget deficits worsened to an average of 2.98 percent in an election year. The rise in the deficit was driven by the decrease in tax collection, which fell by 6.3 percent as spending only rose by percent. Fiscal balances improved after Fernando de la Rua assumed the presidency in 1999, with an average fiscal deficit of 1.20 percent of GDP between 2000 and Additionally, fiscal Barberia and Avelino are with the Getúlio Vargas Foundation. We thank Jorge Streb, Marcela Eslava, two anonymous referees, and especially Ernesto Stein for insights and comments. 1. Drazen (2001); Franzese (2002). 2. Block (2002b); Brender and Drazen (2005c). 101
3 102 ECONOMIA, Spring 2011 balances did not deteriorate during elections, but rather improved slightly. When Néstor Carlos Kirchner was elected president in 2003, Argentina ran a slight fiscal surplus of 0.12 percent. This paper seeks to verify whether these patterns hold systematically for Latin America by exploring two questions. First, are elections catalysts for fiscal policy performance in Latin America? Second, are electoral competitions more likely to provoke larger increases in fiscal deficits during democratic transitions? To answer these questions robustly, we test for political deficit cycles in Latin American democracies employing different measures of democracy, transitions, and election cycles. Our results confirm that elections provoke increases in the fiscal deficit for Latin American democracies, but this pattern is not contingent on a country being in the early phase of its democratic transition. This highlights the importance of the selection criteria used to define democracy and competitive elections when testing for political budget cycles. The paper is structured in the following way. The next section reviews existing theory on the behavior of democracies with respect to government spending, revenue collection, and budget deficits, as well as findings that might clarify why competitive elections held in Latin America during transitional democratic periods may prove to be particularly important and distinct. The paper then describes the time-series cross-sectional data set employed for hypothesis testing and introduces the measures used to test the impact of elections in all Latin American democracies and whether cycles differ when a democratic regime is in a transitional stage. In this section, we also discuss different measures for democracy and recent democratization and the importance of using these measures to undertake more robust testing of the findings reported in earlier studies on political budget cycles in recent democracies. The next section introduces the model specifications used for hypothesis testing in this paper and the battery of alternative models adopted to check the findings for robustness. We then present and discuss the results of the empirical analysis. The final section concludes the paper with a summary of the key findings. Review of the Literature A crucial assumption of political business cycle models is that voters choose leaders on the basis of economic variables, so the degree, nature, and timing of economic policies influence citizens decisions at the ballot box. The electoral
4 Lorena G. Barberia and George Avelino 103 motivations that may guide government policies were discussed by Schumpeter in his study of business cycles and described by Kalecki, but the theoretical framework to describe the opportunistic (office-seeking) motivations of politicians was formally developed much later, by Nordhaus and Tufte. 3 These early and subsequent models are based on the same assumptions: elected leaders in control of monetary policy were able to successfully manipulate economic activity by surprising myopic voters, who were limited to basing their opinions solely on past incumbent performance and inflation rates. More recent theories have made important advances in two important realms. 4 First, models have incorporated forward-looking, rational expectations. 5 Second, research has explored the effects of right- and left-wing party orientation on macroeconomic outcomes during and after elections. 6 Arguing that monetary surprises are an unconvincing driving force for political business cycles, a group of studies has reinvigorated efforts to develop and test models that emphasize fiscal policy as the motivating force for opportunistic cycles. 7 The basic rationale behind models that emphasize the political budget cycle is that governments will manipulate fiscal policy, in part, to obtain electoral success. Based on empirical research, a significant number of recent studies argue that political budget cycles are more acute and more marked in the case of less developed countries. 8 Given the widespread political instability and macroeconomic fluctuations that have marked the region in recent decades, a significant share of the literature on political budget cycles in developing democracies has focused on Latin America. 9 Thus far, the evidence of the electioneering of government expenditures and fiscal balances in Latin America, however, has yielded inconclusive findings. 3. Schumpeter (1939); Kalecki (1943); Nordhaus (1975); and Tufte (1978). 4. Drazen (2001). 5. For example, Rogoff and Sibert (1988); and Rogoff (1990). 6. Alesina (1987). Franzese (2002) provides a valuable summary of the findings emerging from partisan cycles. The effects of ideological orientation on the findings reported in this paper will be a task for future research. 7. Drazen (2001). For a critique and dissenting view of political budget cycles, see Alt and Chrystal (1981). 8. Block (2002a, 2002b); Block, Ferree, and Singh (2003); Shi and Svensson (2002, 2006); Schuknecht (2000). In contrast, Persson and Tabellini (2003) find that political budget cycles are also present in developed democracies in a sample that also includes developing countries, 9. In this section, we have chosen to concentrate our analysis on a discussion emphasizing cross-national empirical research. There have been a number of notable contributions to the study of political budget cycles in specific Latin American countries, including Drazen and Eslava (2010).
5 104 ECONOMIA, Spring 2011 Ames finds that government expenditures rose before and after the sixtyfive elections that took place in seventeen Latin American countries between 1948 and 1970, although only post-election spending proved to be statistically significant. 10 In a later work based on the same group of countries, Ames reports that government expenditures increased by 6.3 percent in the year before and decreased by 7.6 percent in the year after the eighty-two elections that took place between 1947 and In a study of eight South American democracies during the 1980s, Remmer reports that the quarterly percentage change in the fiscal balance is heterogeneous across countries. 12 She argues that elections in Latin America in the 1980s provided leaders with greater political capital to enact reform, given voters preferences for reduced income volatility and inflation. Underscoring the importance of the macroeconomic context in the region, she posits that there is evidence of an anti-political business cycle in presidential elections in Argentina (1989), Bolivia (1985), Brazil (1989), Ecuador (1984 and 1988), Peru (1990), and Venezuela (1988) for the exchange rate and inflation. 13 However, budget deficits were only reduced following the election of Carlos Andrés Pérez Rodríguez in Venezuela, while the election of Carlos Saúl Menem in Argentina was followed by fiscal expansion. More recent studies on the political determinants of government spending and budget deficits in Latin America provide stronger evidence of political budget cycles, but the reported findings are based on only a subset of all Latin American democracies. 14 In a study of eight Latin American countries between 1983 and 1998, Mejía Acosta and Coppedge control for a multiplicity of political determinants and find that budget deficits worsen during elections, though government expenditures do not increase. 15 Their findings are confirmed by Amorim Neto and Borsani, who analyze the influence of presidential and cabinet effects in ten Latin American countries between 1980 and The authors argue that fiscal difficulties during elections are driven by the reluctance of governments to increase taxes. In our paper, we use data from the entire sample of eighteen Latin American democracies for the period 10. Ames (1977). 11. Ames (1987). 12. Remmer (1993). 13. There is robust evidence that the timing of devaluations in Latin America is influenced by the election cycle. See Stein and Streb (1998, 2004) and Stein, Streb and Ghezzi (2005). 14. Amorim Neto and Borsani (2004); Mejía Acosta and Coppedge (2001). 15. Mejía Acosta and Coppedge (2001). 16. Amorim Neto and Borsani (2004),
6 Lorena G. Barberia and George Avelino 105 between 1973 and 2008 to elucidate these earlier suggestive, albeit inconclusive, findings. One of the problems that pose the greatest challenge for interpreting the empirical evidence produced to date is that all too often insufficient attention is given to the selection criteria used to define democracy and competitive elections. This point is underscored by Brender and Drazen, who conclude that if the political budget cycle reflects the manipulation of fiscal policy to improve an incumbent s reelection chances, then it only makes sense in countries in which elections are competitive. 17 Our review of the studies that test for political budget cycles during elections in developing countries reveals that many include contests held under both democratic and authoritarian regimes. 18 For example, Block finds a marked increase in public expenditures on current consumption goods and a decrease in public investment in presidential election years in sixty-nine developing countries between 1975 and However, this study includes both multi-party and single-party elections, thus confounding interpretation as to exactly how regime type might be influencing the reported results. The same problem is found in research specific to Latin America. For example, both of the studies by Ames include elections during periods in which countries were ruled by the military. 20 Mejía Acosta and Coppedge include Mexican presidential elections in which electoral victories were dominated by the Partido Revolucionario Institucional (PRI) and compare these elections with the outcomes from decisions in Argentina, Brazil, Chile, Ecuador, Venezuela, and Uruguay after democracy had returned to these countries. 21 To address this concern, we carefully considered how to define democratic elections before we adopted a research design to test for political budget cycles. An often-cited argument as to why we should expect to find greater political budget cycle effects in developing democracies focuses on the level of development of their political institutions. 22 These studies argue that the dynamics of political competition are very distinct in recent democracies based on both the 17. Brender and Drazen (2005c, p. 1274), 18. Block (2002a, 2002b); Block, Ferree, and Singh (2003); Shi and Svensson (2002, 2006). 19. Block (2002a). 20. Ames (1977, 1987). 21. Mejía Acosta and Coppedge (2001). The authors also recognize this problem stating, although Mexico was not clearly democratic until the 2000 presidential election, it offers a useful example of fiscal performance in a hegemonic party system (p. 9). 22. Persson and Tabellini (2003); Keefer (2005); Keefer and Khemani (2005); Brender and Drazen (2005c); González (2002).
7 106 ECONOMIA, Spring 2011 experience level of voters and the maturity of political institutions. Because voters lack the experience and information to hold elected officials accountable in democracies that have recently transitioned from authoritarian rule, they are more apt to believe campaign promises and can therefore be more easily manipulated by politicians in the first few elections. In addition, political institutions such as the legislature, the judiciary, the central bank, and the media may not be autonomous or institutionalized in the early stages of democracy. 23 In a related work on democratization, Przeworski argues that pressures to increase representation are largely driven by Keynesian coalitions that demand greater redistribution. 24 Accordingly, he argues that incoming elected governments during democratic transitions face a huge backlog of unfulfilled demands, which weakens their ability to effectively manage the economy. Based on the recognized confluence of economic and political crises that usually precipitate democratic transitions, some scholars argue that newly elected governments need to adopt policies that are unsustainable in the medium to long run given the threat of a reversion to autocracy. 25 Brender and Drazen robustly test for the impact of recent democratization on political budget cycles in a cross-section of developed and developing countries from 1960 to They argue that the pattern of political budget cycles in a large cross-section of countries is driven by new democracies and that fiscal manipulation is not statistically significant for established democracies once the sample is appropriately separated. Brender and Drazen find that there is a significant political deficit cycle for new democracies (defined as the first four competitive elections) and argue that higher election-year expenditures in the first few elections are the lever triggering this effect. 27 The authors believe that their findings resonate with earlier studies indicating that voters in developed economies are fiscal conservatives and often tend to 23. Schuknecht (2000). 24. Przeworski (1991). 25. Haggard and Kaufman (1989). 26. Brender and Drazen (2005c). This is not the only criterion that has been used to test for differences between new and established democratic periods based on a specific time period. Rodrik and Wacziarg (2005, p. 51) define a new democracy as a variable that takes on a value of 1 in the year(s) and subsequent five years of any major democratization (as defined by Polity IV), unless the process is interrupted by another major regime change, in which case the dummy is coded as 1 until the interruption. An established democracy is an indicator variable coded 1 for the sixth and subsequent years of a democracy. 27. In contrast to the findings we report for Latin America in this paper, Brender and Drazen (2005c) do not find evidence of a political tax revenue cycle.
8 Lorena G. Barberia and George Avelino 107 remove deficit-producing incumbents from office. 28 They argue that voters in new democracies are less experienced with electoral economics and verify that the net result is higher spending and deficits associated with the first few elections after transition. In a related vein, Block, Ferree, and Singh test whether economic policy performance improves in the early period of democratization. 29 The authors focus on so-called founding elections in sub-saharan African countries between 1980 and 1995, which they define as the first competitive election in which the position of the head of office was openly contested. The authors argue that these elections may be particularly vulnerable to political budget cycles because authoritarian incumbents have much discretion to manipulate expenditures prior to elections. Moreover, nondemocratic leaders who are reluctantly holding elections may also dig deep into government coffers to scare off the opposition, so the winners will undoubtedly have to undertake painful stabilization measures. Countries that have only recently undergone democratization may have reduced capacities to check and balance the powers of the executive branch. Voters may also be more credulous, thus expanding the power of nondemocratic rulers to manipulate fiscal and monetary policies. Based on sixty-five presidential elections, the authors report that competitive multiparty elections (thirty-three of the total elections) are associated with higher monetary growth and government consumption as a share of GDP than noncompetitive elections. However, the hypothesis that founding elections have an additional effect on government spending is not validated. In Latin America, some suggestive evidence that increased political competition during the transition to democracy fuels political budget cycles is provided by González, who studies autocratic Mexican presidential elections between 1957 and She measures increased levels of democratization during elections through lower scores on the Index of Political Coercion and the Autocracy Index. Based on this questionable measure, she argues that greater levels of democracy exacerbated political budget cycles because the PRI responded to the growing threat of losing power by spending more and more resources on election campaigns to ensure its victory. Given that the development of Mexico s political institutions is leading to improvements in transparency and accountability, the study concludes by warning that the 28. Alesina and others (1998); Peltzman (1992). 29. Block, Ferree, and Singh (2003). 30. González (2002).
9 108 ECONOMIA, Spring 2011 election effect will increase as the country becomes more accountable and democratic. The majority of countries in Latin America experienced a founding election marked by the participation of formerly banned political parties and the retreat of the military between the late 1970s and the 1990s. Latin America thus represents an extremely relevant region for examining the vulnerability of democratizing countries to political budget cycles. 31 Specifically, transitions to democracy occurred in Argentina, Bolivia, Brazil, Chile, the Dominican Republic, Ecuador, El Salvador, Guatemala, Peru, and Uruguay between 1974 and Subsequently, democracy also returned to Panama and Paraguay in the mid-1990s and Mexico in One of the striking features of the recent empirical research aimed at testing whether recent democratization is associated with decreases in tax collection and increases in fiscal deficits and government spending is that the beginning and end of the democratic transition is defined as a given number of elections (for example, the first election or the first four competitive elections). Alternative definitions for the period of democratic transition have been developed in political science, but they have not yet been used in empirical political budget cycle research. In this paper we test whether the results of studies that argue that new democracies or democratizing countries are susceptible to political budget cycles differ when these theoretically driven criteria for democratic transitions are employed. Data The data we employ to test for political budget cycles are drawn from a variety of sources. We describe them below. Fiscal Data The dependent variables in this paper are drawn from annual data on central government total expenditure, total revenue and grants, and balance from Government Finance Statistics (GFS) published by the International Monetary Fund (IMF). 32 We used the data set based on the IMF s GFS database that was revised by Brender and Drazen for 1973 to 2001 and added observations for 31. Huntington (1991). 32. IMF (2009).
10 2002 to 2008 for countries in which data were available from the IMF. 33 All three variables are defined in relative terms as a percentage of GDP, which is based on the figures reported by the IMF in its International Finance Statistics (IFS). In all eighteen countries, the fiscal calendar year follows the calendar year. We use the terms fiscal balance and deficit interchangeably in the paper, as most countries ran persistent budget deficits throughout the period. A positive value of the fiscal balance should be interpreted as a budget surplus. Democracy and Election Data Lorena G. Barberia and George Avelino 109 One of the main motivations of this paper is to test the argument that new democracies are more susceptible to political budget cycles, which we do by testing whether study results differ when different criteria for democracy and democratic transitions are used. To explore these differences, we run all empirical tests using two different definitions for democracy and democratic transitions. We restrict the sample to include only democratic years, and for this we employ two definitions of democracy. The first measure is based on Polity IV, analogous both to Brender and Drazen and to Persson and Tabellini; it restricts the sample to countries that received a score between 0 and 10 on the political regime scale, which ranges from 10 (autocracy) to 10 (the highest level of democracy). 34 This is the standard measure used in most empirical research on political budget cycles. The data are based on country assessments by academics using the available literature. 35 However, the measure is subjective and thus is not easily reproducible. The second measure is based on a dichotomous definition of democracy. In contrast to Polity IV, this measure derives from a conceptual definition of democracy, and it is built on empirical observables rather than subjective judgments. The measure is based on minimalist criteria, which we describe as such throughout the paper. 36 The minimalist criterion for the democracy 33. Brender and Drazen (2005c). The raw GFS data was supplemented by IFS data by Brender and Drazen. The procedures are described in Brender and Drazen (2005b). The data set is available online at drazen/. A dummy variable was included in all model estimations to code for the observations added for the period The results reported in this paper are also robust if the data are restricted to Brender and Drazen (2005c); and Persson and Tabellini (2005). 35. Marshall, Jaggers, and Gurr (2008). 36. Democracy is defined simply as the regime in which government offices are filled through competitive elections. See Álvarez and others (1996) for more details.
11 110 ECONOMIA, Spring 2011 dummy variable is drawn from an updated database developed by Cheibub, Gandhi, and Vreeland, which extends the dataset first published in Álvarez, and others in terms of both coverage (country and year) and variables. 37 Under either Polity IV or the Cheibub and others data set, Colombia, Costa Rica, and Venezuela are considered democratic during the entire period. In each case, however, the sample is unbalanced in that the other fifteen countries enter the sample in only some years. In some cases, the Polity IV measure and the minimalist definition are in agreement on the period of democracy. For example, Argentina is excluded in both samples between 1976 and In other cases, there are major differences in the year of entry or exit of a particular country. The most important differences between Polity IV and the minimalist criterion are in the treatment of Mexico and Peru. Based on Polity IV, Mexico is included as a democracy starting in 1988 with the election of Carlos de Salinas de Gortari. In contrast, Cheibub, Gandhi, and Vreeland consider Mexico to be a democracy after the 2000 election in which Vicente Fox assumed power after the Partido Revolucionario Institucional (PRI) which had dominated Mexican presidential elections since 1910 ceded power to the Partido Acción Nacional (PAN). 38 In the case of Peru, the period between the election in 1990, after which Alberto Fujimori staged a coup d état in April 1992, and Fujimbri s resignation in 2000 is considered autocratic by Cheibub, Gandhi and Vreeland, but it is considered part of an uninterrupted period of democracy that began in 1980 if the Polity IV data set is used. As a result of these differences and additional ones for a smaller number of years in the cases of Bolivia, Chile, the Dominican Republic, Ecuador, Guatemala, and Nicaragua, there are forty-six years that are coded differently depending on which measure is used. To replicate the models tested by Brender and Drazen, we used the rule of the year to create a dichotomous dummy variable to code the election year period. 39 Using this rule, we assigned a value of one if an election occurred during the year in question and zero otherwise. 40 Election data are drawn from Nohlen and the Political Database of the Americas (PDBA). 41 Table 1 presents 37. Cheibub, Gandhi, and Vreeland (2010); Álvarez and others (1996). 38. Cheibub, Gandhi, and Vreeland (2010). 39. Brender and Drazen (2005c). 40. We also tested an alternative measure based on the rule of the semester. According to this rule, if an election was held during the first half of year t, then the election year is coded as the year before, or t 1. The results from this alternative method provide a check on the results reported in this paper and are presented in the appendix. 41. Nohlen (2005). The PDBA is maintained by the Center for Latin American Studies at Georgetown University; we used data accessed in 2009.
12 TABLE 1. Presidential Elections in Latin America, Lorena G. Barberia and George Avelino 111 Elections excluded Elections excluded by minimalist by Polity IV Country Presidential election dates criterion criterion Argentina 9/1973, 10/1983, 5/1989, 5/1995, 10/1999, 4/2003, 10/2007 None None Bolivia 6/1980, 7/1985, 5/1989, 6/1993, 6/1997, 6/2002, 12/2005 6/1980 None Brazil 1/1985, 11/1989, 10/1994, 10/1998, 10/2002, 10/2006 None 1/1985 Chile 12/1989,12/1993, 12/1999, 12/ /1989 None Colombia a 4/1974, 4/1978, 5/1982, 5/1986, 5/1990, 6/1994, 6/1998, None None 5/2002, 5/2006 Costa Rica a 2/1974, 2/1978, 2/1982, 2/1986, 2/1990, 2/1994, 2/1998, None None 2/2002, 2/2006 Dominican 5/1974, 5/1978, 5/1982, 5/1986, 5/1990, 5/1994, 6/1996, None 5/1974 Republic 5/2000, 5/2004, 5/2008 Ecuador 4/1979, 1/1984, 1/1988, 7/1992, 7/1996, 6/1998, None None 10/2002, 10/2006 El Salvador 3/1984, 3/1989, 4/1994, 3/1999, 3/2004 None None Guatemala 3/1974, 3/1978, 11/1985, 11/1990, 11/1995, 11/1999, 11/1985 3/1974, 3/1978, 11/2003, 9/ /1985 Honduras 11/1981, 11/1985, 11/1989, 11/1993, 11/1997, 11/2001, 11/1981 None 11/2005 Mexico 7/1988, 8/1994, 7/2000, 7/2006 7/1988, 8/1994 None Nicaragua 11/1984, 2/1990, 10/1996, 11/2001, 11/2006 None 11/1984 Panama 12/1989, 5/1994, 5/1999, 5/2004 None None Paraguay 3/1989, 5/1993, 5/1998, 4/2003, 4/2008 None None Peru 5/1980, 4/1985, 4/1990, 4/1995, 4/2001, 4/2006 4/1990, 4/1995 None Uruguay 11/1984, 11/1989, 11/1994, 10/1999, 10/ /1984 None Venezuela a 12/1973, 12/1978, 12/1983, 12/1988, 12/1993, 12/1998, None None 7/2000, 12/2006 a. No democratic transition. a summary of the elections that are included and specifies which elections are points of contention depending on whether the Cheibub, Gandhi, and Vreeland database or Polity IV are used to define democratic regimes. Using either the Polity IV or minimalist criterion, there are 108 presidential elections in the data set. All eighteen countries had at least one presidential election, but different elections are included or excluded depending on the criterion adopted. On average, there are 6.5 presidential elections per country. We focus on presidential elections and do not include midterm legislative elections. This approach follows the literature; studies that include congressional elections do not find that these elections have distinct impacts on electoral cycles Drazen (2001).
13 112 ECONOMIA, Spring 2011 We also created an interactive term to code elections in which the incumbent president could be reelected. Reelection is prohibited in most Latin American countries; only five countries in the sample permitted immediate reelection during specific periods between 1973 and Incumbents won in eleven of these sixteen elections. The cases of successful immediate reelection are Argentina (Carlos Saúl Menem in 1995), Brazil (Fernando Henrique Cardoso in 1998 and Luiz Inácio Lula da Silva in 2006), Colombia (Álvaro Uribe in 2006), the Dominican Republic (Joaquín Balaguer in 1974, 1986, 1990, and 1994 and Leonel Fernández in 2008), and Venezuela (Hugo Chávez in 2000 and 2006). The dummy variable was coded to equal one if reelection of the president was possible and zero otherwise. As for the coding of democracy, we also test if the results suggesting that there are differences in the behavior of political budget cycles in young democracies are influenced by whether an ad hoc criterion is used. The first measure follows the definition adopted by Brender and Drazen, in which observations for the first four competitive elections are defined as occurring in a new democracy. According to this criterion, 60 of the 108 elections represent new democratic elections. 44 The second measure for recent democratization is based on theoretical underpinnings rather than an arbitrary time period. We created a dichotomous dummy variable that codes one for the democratic transition period. For this measure, the beginning of democratic transition is defined as the year of the inauguration of the first democratic regime following a period of authoritarian rule. 45 Huntington defends the alternation in power of opposition parties as an important criterion for defining the consolidation of democracy. 46 He defines the onset of stable democracy as the second consecutive democratic turnover in which there is a change in the political party controlling the presidency. This two-turnover test, in his opinion, is an unambiguous measure of the resilience of democracy. This definition is also consistent with the definition of democracy adopted in this study following Álvarez and others, who argue that this regime is characterized by the opposition rising to power through elections Payne and others (2007). 44. Brender and Drazen (2005c). 45. O Donnell and Schmitter (1986). 46. Huntington (1991). 47. Álvarez and others (1996).
14 TABLE 2. Lorena G. Barberia and George Avelino 113 Summary Statistics for Democracy and Elections in Latin America, a Polity IV criterion (b) Minimalist criterion Number of democratic years Number of democratic transition years Number of non transition years Percentage of democratic transition years Number of elections (year rule) Number of elections in democratic transition period Percentage of elections in democratic transition period a. The Polity IV criterion for a new democracy is the first four consecutive elections following the end of autocratic rule. The minimalist criterion for a transitional democracy is the founding election, with the democratic transition phase lasting through two election turnovers. According to the two-turnover test, the democratic transition period is coded as the years between the year of the founding election and the year of the second election in which an opposition party wins and assumes office. In the case of Argentina, for example, this implies that the transitional period is between 1983 and 1999 (the year Fernando de la Rua was elected to the presidency, returning the Unión Cívica Radical to power). The year after the de la Rua election is defined as the first year of post-transition democracy. As table 2 summarizes, fifty-three elections in the sample meet these criteria. One of the key challenges in the analysis of elections is the extent to which they may be endogenous, as the end of a particular regime is sometimes not predetermined, but coincides with economic crises. 48 There are a few reasons why the endogeneity of elections does not seem to be a significant problem for the questions explored in this paper. First, the problem of simultaneity bias is much more severe in political budget cycle studies that employ economic growth, unemployment, and inflation as dependent variables, since declines in the performance of these variables are precisely what tend to trigger the collapse of particular administrations. Second, unlike parliamentary democracies, elections are typically held on a fixed schedule in presidential democracies such as those found in Latin America. Of course, there are some notable exceptions. In Bolivia, Hernán Siles Zuazo held presidential elections one year earlier than the end of his term in 1985 in response to rampant hyperinflation. In Argentina, Raúl Alfonsín similarly ceded power earlier than anticipated, though only by a few months. In 2001 Fernando de la Rua resigned from the Argentine presidency after only two years in office in light of massive protests and a spiraling economic crisis. 48. Przeworski and Limongi (1993); Haggard and Kaufman (1997).
15 114 ECONOMIA, Spring 2011 Control Variables The demographic and economic variables included as control variables are analogous to those used by Brender and Drazen. 49 Demographic characteristics of the population are likely to influence government spending. Two demographic variables representing the fraction of the population aged fifteen to sixty four and aged sixty-five and over are used as controls. A higher percentage of elderly and young people in the population are expected to positively increase budget allocations for social programs and social security, leading to increases in fiscal spending and the worsening of budget deficits. Unless otherwise noted, the control variables data are from World Development Indicators. 50 Given the heterogeneity in income and growth rates across the region, it is important to include economic controls. The first is the level of economic development, defined as the real gross domestic product per capita and measured in constant 2000 U.S. dollars. Per capita income is included in the model to control for Wagner s law, which holds that the level of public spending is positively correlated with levels of economic development. Higher levels of per capita income are expected to be correlated with higher levels of government spending. A control for the output gap or the proportion of growth that is unexpected in a given year is also included, again following Brender and Drazen. 51 We use the log-difference between real GDP and its (country-specific) trend (computed using the Hodrick-Prescott filter). Trade liberalization increased dramatically in Latin America in the 1990s. We thus include a measure of trade integration to control for the degree of an economy s integration with world markets. Trade is calculated as the sum of imports and exports relative to GDP, where the denominator is calculated by converting domestic local currency to current U.S. dollars based on exchange rate conversions. In addition to these variables, a dummy variable was included to control for the additional years that were added to the Brender and Drazen data set from the IMF s GFS database. Estimation Procedure and Model Specification The baseline model used to test the effect of elections on fiscal variables is as follows: 49. Brender and Drazen (2005c). 50. World Bank (2009). 51. Brender and Drazen (2005c).
16 Lorena G. Barberia and George Avelino 115 () 1 Y ˆ ˆ ˆ it, = αt + β1yit, 1 + β2zit, + β3elecit, + ci + μ it The three measures of fiscal policy, which are analogous to those used by Brender and Drazen, are total government spending as a share of GDP, total revenue collection as a share of GDP, and the budget balance as a share of GDP. 52 Z is a vector of control variables as described earlier, and α represents year dummies. The index i refers to the N observational units (or panels), and t indexes the T time periods. The term c i is a dummy variable for each country, intended to capture unobserved country-specific effects, while µ it is an error term associated with unit i at time t. This model follows the literature and tests whether there are differences in spending prior to elections by including a dummy variable, ELEC, for the election year. We check the robustness of political budget cycles to alternative definitions of democracy by employing the two different definitions described in the previous section of this paper. Specifically, we code a given election year as one if a country received a score indicating that it was a democracy based on the Polity IV criteria, and we then compare these results with those obtained when democratic elections are determined using the minimalist criterion defined by Cheibub, Gandhi, and Vreeland. 53 We seek to verify if both measures yield a positive coefficient that is significantly different from zero in the year of the election. Based on the assumption that past fiscal policy levels influence future levels, we include a lagged dependent variable in each specification. We used a series of alternative estimation strategies to check for consistency and robustness of the results; these are reported in the next section of this paper. First, we estimated pooled ordinary least squares (OLS) regressions with panel corrected standard errors (column 1). 54 We also estimated country fixed effects (column 2) and year and country fixed effects (column 3). 55 In addition, two,. 52. Brender and Drazen (2005c). 53. Cheibub, Gandhi, and Vreeland (2010). 54. The model was estimated with the Stata XTPCSE command. 55. For fixed T, Nickell (1981) demonstrates that the within-group estimate of the coefficient is likely to be biased downward by a factor of 1/T, where T is the length of the panel. Thus, the magnitude of the bias in the fixed-effects estimates can be calculated for the within-group estimator for a dynamic model with fixed individual effects. The exact magnitude depends on which sample and indicator are used, as some countries do not report data for the entire period. In a panel of all countries from 1973 to 2008, the length of the sample ranges from thirty-six years to a minimum length of nineteen years for three countries (Honduras, Nicaragua, and Paraguay). Hence, the bias from using a fixed-effects estimator in these regressions is likely to range from 2.77 percent (1/36) to 5.26 percent (1/19).
17 116 ECONOMIA, Spring 2011 generalized method of moments (GMM) procedures were used: Arellano and Bond s first-differenced GMM estimator and Blundell and Bond s system GMM estimator (columns 4 and 5). 56 Therefore, tables 3 through 6 consist of five columns. The GMM difference and systems equations use two lags of the dependent variable. The GMM estimates use the Arellano-Bond (difference) and Blundell-Bond (system) procedures with orthogonal deviations to adjust for an unbalanced panel; the deviations were collapsed to minimize the number of instruments following the recommendations of Roodman. 57 Per capita GDP and the log difference between real GDP and its (country-specific) trend (computed using the Hodrick-Prescott filter) were also included as endogenous variables in the GMM estimations. For GMM estimates, standard errors are reported as t statistics based on Windmeijer finite sample correction and corrected for serial correlation and heteroskedasticity. 58 To test whether the election effect depends on an election taking place under either a new democracy or a transitional democracy, we undertake a second estimation: ( ) Y = α + βˆ Y + βˆ Z + βˆ ELEC + βˆ RECENTDEM 2 it, t 1 it, 1 2 it, 3 it, 4 βˆ + ( ) ELEC RECENTDEM c it i μ,, i, t where ELEC and RECENTDEM are dummy variables coding years considered election and democratic transition years. The variable ELEC RECENTDEM is an interactive variable equal to one if the presidential election took place during the transitional period and zero otherwise. The marginal impact of the recent democratization period on fiscal performance is captured by β $ 4. The marginal effect of an election during the democratic transition phase is captured by β $ 5. In the next section of the paper, we present the results of the test carried out to verify whether fiscal spending and deficits increase during elections in recent democracies. We do so by examining if β $ 3 +β $ 5 (the total marginal it, 56. Arellano and Bond (1991); Blundell and Bond (1998). The exercise and commands for GMM estimation are based on Roodman (2006) and were carried out using Stata We also carried out GMM estimates controlling for year fixed effects. The results coincided with the GMM results without controls for year fixed effects. 57. Roodman (2006). 58. Windmeijer (2005). We also tested an error correction model (ECM), which is also appropriate for highly persistent series, with panel corrected standard errors based on the first difference of the dependent variables; our findings did not change.
18 Lorena G. Barberia and George Avelino 117 effect of an election in a recent democracy) is significantly different from zero. 59 We also further test the robustness of political budget cycles to alternative definitions of transitional democracy. Our intention here is to focus on whether the observational criteria of democratic transitions produce results that are different from measures that rely on subjective operational rules. Results This section summarizes the findings from tests on political budget cycles in Latin American democracies in the nearly four decades between 1973 and Regardless of whether Polity IV or a minimalist criterion is adopted for democracy, the results suggest that there are political budget cycles in Latin American democracies. In this respect, our tests confirm Brender and Drazen s findings. 61 However, our results show that arguments that these cycles are driven by recently democratized countries are less robust. Specifically, we find that evidence of political budget cycles in recent democracies is dependent on the definition of a transitional democracy. As we report below, the results produced when models are estimated based on Polity IV definitions of democracy and new democracies are not confirmed when we use the minimalist criterion of two turnovers of political power to define the transition period. Tables 3 and 4 examine whether there are troughs in revenue collection or peaks in government spending and the fiscal deficit in the year of a presidential election in Latin America. As observed earlier, all of the dependent variables are measured as a share of GDP. For presentation purposes, only the estimate of the coefficient for the dummy variable for the election is presented. 62 The base group is all other democratic years. Table 3 presents the results of the five specifications in which the election year was only coded as valid if the country was judged a democracy based on Polity IV. Table 4 presents the results of the same test based on the minimalist definition of democracy. The results in both tables suggest that there are important increases in fiscal deficits (panel C) in Latin America. The fiscal balance can be either negative or positive; a negative coefficient indicates a worsening of the government s 59. Braumoeller (2004); Brambor, Clark, and Golder (2006). 60. We also carried out the same tests using the original Brender and Drazen (2005c) data set for the period between 1973 and The results of that exercise not only confirmed the findings reported here, but generally were stronger in terms of statistical significance. 61. Brender and Drazen (2005c). 62. The complete results are available on request.
19 TABLE 3. The Effect of Elections on Political Budget Cycles in Latin America, : Polity IV Definition a Pooled OLS Pooled OLS with GMM GMM with country country and year one-step first one-step Pooled OLS fixed effects fixed effects difference system Dependent and explanatory variable (1) (2) (3) (4) (5) A. Government spending/gdp Election year (ELEC t ) (0.280) (0.263) (0.272) (0.299) (0.310) No. observations Avg. time series length R squared No. instruments 9 17 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system B. Government revenue/gdp Election year (ELEC t ) * (0.296) (0.272) (0.280) (0.355) (0.343) No. observations Avg. time series length R squared No. instruments 9 17 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system C. Fiscal balance/gdp Election year (ELEC t ) 0.633** 0.740** 0.704** 0.776* 0.769* (0.309) (0.296) (0.291) (0.376) (0.422) No. observations Avg. time series length R squared No. instruments 9 17 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system *Statistically significant at the 10 percent level. **Statistically significant at the 5 percent level. a. The covariates include lags of the dependent variable, the log of per capita GDP, the ratio of international trade (sum of merchandise exports and imports) to GDP, the fraction of the population over age sixty-five, the fraction of the population between ages fifteen and sixty-four, and the log-difference between real GDP and its (country-specific) trend, estimated using a Hodrick-Prescott filter. Pooled OLS regressions were estimated with panel corrected standard errors that correct for groupwise heteroskedasticity and contemporaneous correlations of the errors. Where noted, country and year dummy variables were included in the regressions, but they are not reported above for reasons of space. The two GMM specifications estimate the Arellano-Bond (difference) and Blundell-Bond (system) procedures with orthogonal deviations to adjust for an unbalanced panel; they are collapsed to minimize the number of instruments following the recommendations of Roodman (2006). Per capita GDP and growth were also included as endogenous variables in the GMM estimations. Standard errors are in parentheses. For the GMM estimates, standard errors are reported as t statistics based on Windmeijer (2005) finite sample correction and are corrected for serial correlation and heteroskedasticity. b. The Arellano-Bond tests for first-order and second-order serial correlation are on the first-differenced residuals. The p values are the probability of rejecting the null hypothesis of no autocorrelation.
20 TABLE 4. The Effect of Elections on Political Budget Cycles in Latin America, : Minimalist Criterion a Pooled OLS Pooled OLS with GMM GMM with country country and year one-step first one-step Pooled OLS fixed effects fixed effects difference system Dependent and explanatory variable (1) (2) (3) (4) (5) A. Government spending/gdp Election year (ELEC t ) (0.291) (0.271) (0.289) (0.277) (0.319) No. observations Avg. time series length R squared No. instruments 9 17 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system B. Government revenue/gdp Election year (ELEC t ) ** (0.305) (0.279) (0.296) (0.358) (0.346) No. observations Avg. time series length R squared No. instruments 9 17 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system C. Fiscal Balance/GDP Election year (ELEC t ) 0.598* 0.669** 0.709** 0.732* (0.323) (0.305) (0.315) (0.370) (0.404) No. observations Avg. time series length R squared No. instruments 9 14 Arellano-Bond test for AR(2) (p value) b Hansen test for joint validity of Diff. Sargan tests for all system *Statistically significant at the 10 percent level. **Statistically significant at the 5 percent level. a. The covariates include lags of the dependent variable, the log of per capita GDP, the ratio of international trade (sum of merchandise exports and imports) to GDP, the fraction of the population over age sixty-five, the fraction of the population between ages fifteen and sixty-four, and the log-difference between real GDP and its (country-specific) trend, estimated using a Hodrick-Prescott filter. Pooled OLS regressions were estimated with panel corrected standard errors that correct for groupwise heteroskedasticity and contemporaneous correlations of the errors. Where noted, country and year dummy variables were included in the regressions, but they are not reported above for reasons of space. The two GMM specifications estimate the Arellano-Bond (difference) and Blundell-Bond (system) procedures with orthogonal deviations to adjust for an unbalanced panel; they are collapsed to minimize the number of instruments following the recommendations of Roodman (2006). Per capita GDP and growth were also included as endogenous variables in the GMM estimations. Standard errors are in parentheses. For the GMM estimates, standard errors are reported as t statistics based on Windmeijer (2005) finite sample correction and are corrected for serial correlation and heteroskedasticity. b. The Arellano-Bond tests for first-order and second-order serial correlation are on the first-differenced residuals. The p values are the probability of rejecting the null hypothesis of no autocorrelation.